Civil Law and UAE Automated Moral Reasoning Systems in Legal Tech

1. Introduction

Automated moral reasoning systems are legal-technology systems designed to evaluate legal or quasi-legal decisions according to concepts such as fairness, proportionality, good faith, equality, public interest, harm prevention, transparency, accountability, and ethical compliance.

In the UAE, such systems may be used in:

• automated compliance;
• banking and AML monitoring;
• insurance decisions;
• smart contracts;
• dispute-resolution platforms;
• legal-risk scoring;
• corporate governance;
• employment decision-making;
• AI-assisted judicial or administrative analysis;
• fraud detection;
• consumer protection; and
• regulatory technology (RegTech).

The important legal point is that an automated system may assist in moral or legal reasoning, but it does not itself become the source of UAE law. The final legal authority remains the Constitution, legislation, regulations, applicable Sharia principles, established legal principles, contractual obligations within their lawful limits, and the competent court or public authority.

The UAE's current Civil Transactions framework is especially important because Federal Decree-Law No. 25 of 2025 promulgating the Civil Transactions Law came into force on 1 June 2026, replacing the former 1985 Civil Transactions Law. Pre-2026 cases therefore remain useful as historical and doctrinal authorities, but they should not automatically be treated as interpretations of the new Code.

2. Meaning of Automated Moral Reasoning

An automated moral reasoning system can be understood as:

A software or AI system that applies predefined ethical, legal, policy, or value-based criteria to facts and produces a recommendation about what action is fair, permissible, proportionate, or appropriate.

For example, a bank's AI system could evaluate:

Transaction → Risk indicators → Applicable rules → Ethical/legal factors → Recommendation

It might classify a transaction as:

• low risk;
• suspicious;
• requiring human review;
• temporarily restricted; or
• requiring escalation to a compliance officer.

Similarly, an employment platform could assess whether an automated decision produces apparently discriminatory or disproportionate outcomes.

However, moral evaluation and legal adjudication are not identical.

3. Moral Reasoning Versus Legal Reasoning

Therefore, an AI system cannot simply state:

“This outcome is morally unfair, therefore it is legally invalid.”

The system must identify the legal rule that gives the fairness consideration legal significance.

4. Civil-Law Foundation in the UAE

The UAE is fundamentally a codified legal system.

This means that automated legal reasoning should begin with applicable written law rather than treating algorithmic morality as an independent source of law.

The current Civil Transactions Law establishes a hierarchy for situations not expressly resolved by legislation. The system therefore has to distinguish between:

1. mandatory statutory rules;
2. applicable regulations;
3. principles derived from Islamic jurisprudence where legally relevant;
4. custom;
5. principles of natural law and justice where the statutory framework permits their use; and
6. judicial interpretation and application.

Consequently, an AI moral-reasoning engine should not be programmed as though its own ethical preferences have legislative status.

5. Article 1 and Automated Legal Reasoning

The current Civil Transactions Law's methodology is particularly significant.

Where a dispute is expressly or implicitly regulated by legislation, the legislative rule takes priority.

Where no applicable legislative provision exists, the court proceeds through the legally prescribed supplementary sources.

This has an important technological consequence:

AI should operate as a reasoning assistant, not an autonomous lawmaker.

An AI system could therefore be designed to ask:

1. What is the relevant statutory provision?
2. Is the provision mandatory or discretionary?
3. Are there applicable regulations?
4. Is the issue governed by public order?
5. Are contractual terms relevant?
6. Is custom legally relevant?
7. Is good faith relevant?
8. Is proportionality relevant?
9. What factual evidence supports the conclusion?
10. Does the proposed result remain within the authority of the decision-maker?

6. Good Faith and Automated Moral Reasoning

Good faith is particularly important.

Under the former Civil Transactions Law, Article 246 established the principle that contracts must be performed in accordance with their contents and in good faith. The new Civil Transactions Law continues the importance of good-faith contractual performance.

An automated system may therefore be programmed to identify conduct such as:

• deliberate concealment;
• opportunistic contractual behavior;
• inconsistent representations;
• manipulation of contractual mechanisms;
• abusive exercise of contractual rights; or
• conduct inconsistent with legitimate contractual expectations.

But the AI's identification of "bad faith" is only an analytical indicator.

The court must ultimately determine whether the legal requirements for bad faith or abuse have actually been established.

7. Abuse of Rights

Another important foundation is the UAE doctrine of abuse of rights.

The former Article 106 prohibited abusive exercise of rights in specified circumstances, including situations involving disproportionate harm, unlawful purposes, or conduct contrary to accepted legal principles.

The new Civil Transactions Law retains the abuse-of-rights principle.

An automated moral reasoning system could therefore identify:

legitimate right + disproportionate exercise + foreseeable harm = possible abuse-of-rights issue.

For example, a creditor might technically possess a contractual right to take a particular action, but an AI system may flag that exercising the right in an extreme manner could raise proportionality or abuse concerns.

The system should flag the issue, not conclusively adjudicate it.

8. Proportionality as an AI Reasoning Principle

Proportionality is particularly important when automated systems affect individuals.

Consider an automated compliance system that detects unusual banking activity.

There may be several possible responses:

Level 1: additional information request
Level 2: enhanced due diligence
Level 3: human investigation
Level 4: temporary restriction
Level 5: referral to competent authorities

A morally and legally sophisticated system should not automatically select Level 5 merely because its risk score is high.

Instead it should consider:

• seriousness of the suspected conduct;
• quality of evidence;
• reliability of the data;
• possibility of false positives;
• financial impact;
• legitimate explanations;
• urgency;
• statutory authority; and
• availability of less restrictive measures.

This is the difference between risk detection and legally justified enforcement.

9. Human Oversight

Human oversight is one of the most important safeguards.

An automated moral reasoning system should ideally have:

Human-in-the-loop

A human decision-maker reviews the AI recommendation before legal consequences arise.

Human-on-the-loop

The AI operates automatically but remains subject to continuous human supervision.

Human-in-command

The human authority retains ultimate power to override the system.

For high-impact decisions, the third model is especially important.

10. UAE Evidence Law and Automated Reasoning

The UAE Evidence Law, Federal Decree-Law No. 35 of 2022 on Evidence in Civil and Commercial Transactions, gives electronic evidence substantial legal recognition.

Electronic evidence can include:

• electronic records;
• electronic instruments;
• electronic signatures;
• electronic correspondence;
• electronic communications;
• electronic media; and
• other electronically generated or stored evidence.

This is highly relevant to automated moral reasoning.

The existence of an AI-generated recommendation does not automatically establish that its conclusion is correct.

A court may need to examine:

• the underlying data;
• source of the data;
• algorithm;
• system logs;
• methodology;
• integrity;
• authenticity;
• human intervention;
• audit trail; and
• reliability.

11. AI Output Is Not Automatically Conclusive Evidence

An important distinction should be made:

Electronic evidence

The system's electronic record may be admissible.

AI-generated conclusion

The substantive conclusion generated by the AI may still require assessment.

For example:

"AI system determined that the transaction was fraudulent."

This statement alone does not establish fraud.

The court may ask:

• What data did the system use?
• Was the data accurate?
• Was the algorithm functioning correctly?
• What threshold was used?
• Were alternative explanations considered?
• Was the system biased?
• Was the model changed after the event?
• Can the decision be reproduced?
• Did a qualified human review the result?

12. AI Explainability

An automated moral reasoning system should be capable of explaining its decision.

A useful model is:

Input → Rule → Weight → Reasoning → Output → Human Review

For example:

Customer's transaction was classified as high risk because three independently verified indicators exceeded the regulatory threshold.

This is considerably more defensible than:

AI determined that the customer was unethical.

The second formulation is problematic because "unethical" is vague, subjective and potentially discriminatory.

13. Algorithmic Bias

Automated moral reasoning creates substantial risks of bias.

Bias can arise from:

• historical data;
• incomplete data;
• proxy variables;
• geographical classifications;
• socioeconomic indicators;
• language;
• nationality-related variables;
• behavioral assumptions;
• inaccurate risk scoring;
• unequal error rates.

A civil-law analysis therefore needs to ask whether an automated system has produced an unlawful or unjustified disadvantage.

The system should be tested for:

• false positives;
• false negatives;
• disparate outcomes;
• inconsistent treatment;
• data-quality problems;
• unexplained exclusions; and
• discriminatory proxies.

14. Automated Moral Reasoning and Public Order

Some UAE legal rules involve public order.

Where a rule is mandatory and connected with public order, private parties cannot simply contract around it.

This creates an important limitation for legal-tech systems.

An AI system cannot say:

"The parties agreed to waive this rule, therefore the rule does not apply."

The system must first determine whether the relevant rule is capable of contractual modification.

15. Automated Contracting and Smart Contracts

Automated moral reasoning can also be incorporated into smart contracts.

Suppose a smart contract automatically imposes a penalty after a specified event.

The software might conclude:

"Condition satisfied → penalty automatically executed."

But legally relevant questions remain:

• Was the contract valid?
• Did the parties possess capacity?
• Was consent genuine?
• Was the triggering event correctly identified?
• Was there force majeure?
• Was performance prevented?
• Was the clause abusive?
• Is the penalty subject to judicial adjustment?
• Does public order restrict enforcement?

Thus:

Code execution ≠ complete legal validity.

16. Automated Compliance and AML

This is one of the most significant practical applications in the UAE.

The current UAE AML framework under Federal Decree by Law No. 10 of 2025 Regarding Anti-Money Laundering, Combating the Financing of Terrorism and Proliferation Financing permits extensive mechanisms involving:

• suspicious transaction monitoring;
• temporary suspension;
• freezing;
• tracing;
• seizure;
• confiscation;
• asset recovery;
• beneficial ownership analysis; and
• international cooperation.

An automated moral-reasoning engine can assist institutions in identifying suspicious patterns.

But the system should not itself be treated as possessing statutory enforcement authority.

Correct model

AI detection → compliance review → competent authority → legally authorized action

rather than:

AI detection → automatic deprivation of rights

17. Automated Moral Reasoning and Asset Recovery

Asset-recovery systems can use algorithms to identify:

• suspicious transfers;
• beneficial owners;
• related accounts;
• shell entities;
• cryptocurrency movements;
• common directors;
• unusual transaction patterns;
• asset transfers following litigation;
• rapid movement of funds between jurisdictions.

This is particularly useful for tracing complex financial networks.

But tracing is an evidentiary function.

The fact that an algorithm identifies a relationship does not necessarily prove:

• ownership;
• knowledge;
• fraud;
• criminal origin;
• conspiracy; or
• legal responsibility.

Those conclusions require legal analysis and evidence.

18. Six Important UAE Case Laws

Because UAE courts have not yet developed a large body of reported cases specifically concerning AI moral reasoning systems, the following authorities should be understood as foundational or analogical authorities concerning evidence, expert reasoning, contractual good faith, abuse of rights, and automated/technical decision-making principles.

Case 1: UAE Federal Supreme Court, Civil Appeal No. 79/2020

Principle

The Federal Supreme Court dealt with the legal significance of an admission.

The Court emphasized that an admission involves recognition of a right and may relieve the opposing party from the ordinary burden of proving that admitted matter.

Relevance to AI

An AI system may identify an electronically recorded admission, such as:

• email;
• electronic message;
• digital acknowledgment;
• automated statement.

But the legal significance of the statement still depends on:

• authenticity;
• attribution;
• context;
• capacity; and
• applicable evidentiary rules.

Therefore:

AI identification of an admission ≠ automatic legal conclusion.

19. Case 2: UAE Federal Supreme Court, Commercial Appeal No. 215/2020

Principle

The Court addressed reliance on expert evidence and emphasized the importance of a properly reasoned expert report.

A court may rely upon expert analysis where it is properly reasoned, but merely adopting an expert conclusion without addressing material issues can create a reasoning deficiency.

Relevance to AI

This principle is extremely important for AI-generated expert reports.

Suppose an AI system produces:

"Probability of liability: 87%."

That percentage cannot replace legal reasoning.

The decision-maker must understand:

• how the percentage was calculated;
• what evidence was considered;
• what assumptions were made;
• whether competing evidence was considered; and
• why the conclusion is legally relevant.

20. Case 3: UAE Federal Supreme Court, Penal Cassation No. 1422/2022

Principle

The Court emphasized that evidence supporting a judgment must possess sufficient probative value and that evidence should be examined and assessed rather than treated mechanically.

Relevance to automated moral reasoning

An AI system may identify thousands of indicators.

Quantity, however, does not equal quality.

For example:

100 weak indicators do not necessarily establish one legally sufficient fact.

Automated systems therefore need evidentiary weighting rather than merely counting correlations.

21. Case 4: UAE Federal Supreme Court, Penal Cassation No. 660/2023

Principle

The Court recognized the trial court's ability to form its conviction from the material presented before it, provided that the inference is sound and grounded in the evidence.

Relevance to AI

This supports an important distinction:

AI inference can assist human legal inference, but cannot replace the judicial function.

An algorithm may detect:

Transaction A → Account B → Company C → Beneficiary D.

The judge must still determine what legal significance that chain has.

22. Case 5: UAE Federal Supreme Court, Penal Cassation No. 1093/2019

Principle

The Court recognized the trial court's authority to evaluate and weigh evidence and to rely upon evidence it finds reliable and probative.

Relevance

This demonstrates why an AI system should present evidence in a form that permits meaningful evaluation.

A good legal-tech system should therefore provide:

• source documents;
• confidence levels;
• timestamps;
• data provenance;
• contradictory evidence;
• alternative explanations; and
• an audit trail.

An opaque score alone is inadequate for serious legal decision-making.

23. Case 6: Abu Dhabi Court of Cassation, Case No. 55/2016

Principle

The Court considered the doctrine of abuse of rights under the former Civil Transactions Law.

The underlying principle is that the exercise of a formally recognized right can become legally objectionable where it falls within the statutory circumstances constituting abusive exercise.

Relevance to automated moral reasoning

This case is highly relevant conceptually.

An AI system can identify:

"Formal right exists."

But it must also ask:

"Could the manner in which the right is being exercised constitute abuse?"

For example, an automated debt-collection system could repeatedly impose technically authorized measures without considering circumstances that make the conduct disproportionate or abusive.

The AI should therefore function as a risk detector, not an automatic enforcement mechanism.

24. Case 7: Dubai Court of Cassation, Civil Appeal No. 6/2017

Principle

The Court addressed contractual obligations and good-faith performance.

Relevance

Automated contract-management platforms increasingly monitor contractual performance.

The system should therefore evaluate not merely:

"Was a literal contractual condition satisfied?"

but also whether relevant legal principles concerning good faith and lawful contractual performance are engaged.

This is especially important for automated termination, penalties and suspension mechanisms.

25. Case 8: Dubai Court of Cassation, Appeal No. 313/2007

Principle

The case concerned contractual termination and the exercise of contractual powers.

Relevance

An automated legal system may be programmed to trigger termination automatically.

But the existence of a termination clause does not necessarily mean that every automated trigger is legally unchallengeable.

The system should check:

• contractual wording;
• notice requirements;
• applicable mandatory rules;
• good faith;
• factual triggering conditions;
• waiver;
• performance history; and
• consequences of termination.

26. Case 9: Dubai Court of Cassation, Appeal No. 440/2016

Principle

The case is relevant to contractual stability, good faith and the limits surrounding contractual termination.

Relevance

An AI system designed to recommend termination should distinguish between:

mechanical contractual trigger

and

legally justified termination.

That distinction is central to automated legal decision-making.

27. Case 10: UAE Federal Supreme Court, Penal Cassation No. 891/2022

Principle

This was an important UAE AML case concerning the assessment of evidence and money-laundering liability under the then-applicable AML legislation.

The Court recognized the ability of the trial court to assess the factual material before it and addressed the independent nature of money-laundering offences.

Relevance

This is particularly significant for automated compliance technology.

AI may identify suspicious financial patterns, but:

suspicion is not identical to legally established liability.

Automated AML systems must therefore distinguish:

risk score → suspicion → investigation → evidence → legal finding.

The case predates the current 2025 AML legislation and should therefore be used as a foundational AML authority rather than as an interpretation of the current statute.

28. Summary of the Case-Law Principles

29. Automated Moral Reasoning and the Duty to Give Reasons

One of the strongest safeguards is a reason-giving requirement.

An automated legal-tech system should ideally generate an explanation such as:

"The transaction was flagged because four independently verified indicators were present. Two indicators were subsequently excluded because the underlying data could not be authenticated. The remaining indicators justify enhanced review but do not independently establish unlawful conduct."

This is much safer than:

"Risk score: 94/100 — block customer."

The first approach facilitates:

• review;
• challenge;
• correction;
• accountability;
• judicial scrutiny; and
• auditability.

30. Algorithmic Accountability

A UAE legal-tech provider or regulated institution should maintain:

1. Data provenance

Where did the data originate?

2. Model documentation

What methodology does the system use?

3. Version control

Which version of the algorithm produced the result?

4. Audit logs

What did the system do and when?

5. Human review

Who reviewed the output?

6. Error monitoring

How often does the system produce false positives?

7. Override mechanisms

Can an authorized human reverse the decision?

8. Evidence preservation

Can the organization reproduce the decision later?

31. Moral Machine Problems in UAE Legal Technology

Automated moral reasoning becomes especially difficult when two legitimate values conflict.

For example:

Privacy vs security

An institution wants extensive customer monitoring to detect fraud.

Contractual freedom vs fairness

A business wants strict automated enforcement of contractual penalties.

Efficiency vs due process

A regulator wants rapid automated decisions.

Financial crime prevention vs customer rights

A bank wants to prevent suspicious transactions while avoiding unjustified restrictions.

Commercial confidentiality vs transparency

A company wants to protect its algorithm while a litigant seeks to challenge the decision.

The system therefore needs a legal conflict-resolution hierarchy, not merely a morality score.

32. AI and Natural Justice

Where an automated system significantly affects a person's legal or economic position, principles associated with procedural fairness become important.

A robust system should allow, where legally appropriate:

• notice;
• explanation;
• human review;
• correction of inaccurate data;
• challenge of the decision;
• consideration of relevant circumstances;
• preservation of evidence.

The precise procedural rights depend on whether the decision is made by:

• a private company;
• regulated financial institution;
• government authority;
• court;
• arbitral tribunal; or
• another legally empowered body.

33. AI Moral Reasoning in Arbitration

Automated moral reasoning can assist arbitral tribunals with:

• document classification;
• chronology;
• contractual interpretation support;
• damages analysis;
• evidence organization;
• procedural fairness checks;
• conflict identification.

But arbitrators cannot delegate their adjudicative responsibility to AI.

The tribunal must independently determine:

• jurisdiction;
• applicable law;
• facts;
• evidence;
• liability;
• damages; and
• procedural fairness.

This is consistent with the fundamental principle that an arbitral tribunal derives authority from the arbitration agreement and applicable arbitration law—not from an AI system.

34. AI and Expert Evidence

AI systems can operate as sophisticated technical tools.

For example:

AI system → forensic financial analysis → expert → expert report → court

This is more legally defensible than:

AI system → automatic judgment

The human expert can explain:

• methodology;
• assumptions;
• limitations;
• error rates;
• source data;
• competing interpretations.

The court then independently assesses the evidence.

35. AI Hallucination and Legal Reasoning

A major problem is AI hallucination.

An AI system may:

• invent a legal provision;
• misstate a case;
• confuse jurisdictions;
• attribute a principle to the wrong court;
• fabricate evidence;
• incorrectly summarize a contract.

For UAE legal technology, this creates serious risks because the system may appear authoritative while producing legally incorrect reasoning.

Therefore every high-impact legal AI system should have:

source verification + citation validation + human review + auditability.

36. AI Moral Reasoning and Legal Causation

A particularly difficult issue is causation.

Suppose an automated system incorrectly identifies a customer as suspicious.

The customer subsequently suffers:

• account restriction;
• business interruption;
• loss of a transaction;
• reputational harm.

The legal question is not simply:

"Did the algorithm make an error?"

It is:

"Did the legally attributable conduct cause compensable harm under the applicable legal rules?"

The analysis may require:

1. duty;
2. breach;
3. causation;
4. actual damage;
5. foreseeability where relevant;
6. defenses;
7. contributory conduct where legally relevant; and
8. applicable contractual or statutory limitations.

37. Automated Moral Reasoning and Civil Liability

Potentially responsible parties can include:

• AI developer;
• software vendor;
• data provider;
• financial institution;
• employer;
• compliance department;
• system operator;
• human decision-maker.

Responsibility should not automatically be assigned to "the AI."

An AI has no independent legal personality merely because it produces an autonomous output.

The legal analysis must identify the human or legal person legally responsible for deploying, controlling, relying upon, or negligently operating the system.

38. The "Black Box" Problem

A black-box algorithm creates a fundamental legal problem.

If the institution says:

"We cannot explain why the AI reached this result."

the court may have difficulty determining:

• reliability;
• causation;
• authenticity;
• procedural fairness;
• discriminatory effects;
• compliance with mandatory law.

Therefore, high-impact systems should favor explainability over maximum predictive complexity where the additional complexity cannot be justified.

39. Civil-Law Model for UAE Automated Moral Reasoning

A useful UAE legal-tech architecture would be:

Layer 1 — Constitutional limits

↓

Layer 2 — Applicable legislation

↓

Layer 3 — Sector-specific regulation

↓

Layer 4 — Contractual rules

↓

Layer 5 — Good faith and abuse-of-rights principles

↓

Layer 6 — Evidence and factual analysis

↓

Layer 7 — Ethical/proportionality assessment

↓

Layer 8 — AI recommendation

↓

Layer 9 — Human legal review

↓

Layer 10 — Judicial/administrative accountability

This model prevents ethical AI output from becoming a substitute for legal authority.

40. Key Legal Risks

41. Best-Practice Framework for UAE Legal-Tech Developers

A legally responsible automated moral reasoning system should incorporate:

A. Legal hierarchy

The system should identify the governing legal rule before applying ethical considerations.

B. Explainability

Every material recommendation should have an understandable rationale.

C. Evidence provenance

The system should identify the source of important information.

D. Human oversight

High-impact decisions should receive qualified human review.

E. Proportionality

The system should recommend the least excessive legally available intervention where appropriate.

F. Bias testing

Regular statistical and legal testing should be conducted.

G. Auditability

Every important decision should produce a permanent audit trail.

H. Contestability

Affected persons should have legally appropriate avenues for challenge.

I. Data protection

Personal information should be processed consistently with applicable UAE data-protection requirements.

J. Continuous monitoring

The system should be re-evaluated when laws, regulations, datasets or algorithms change.

42. Difference Between Automated Compliance and Automated Moral Reasoning

The two systems can be integrated, but they should not be confused.

43. Central Legal Principle

The most important principle can be expressed as:

An automated system may automate legal analysis, but it cannot automatically manufacture legal authority.

Similarly:

An algorithm may identify a morally problematic result, but only applicable law can determine its legal consequences.

This distinction is fundamental to UAE civil-law technology.

44. Conclusion

Automated moral reasoning systems in UAE legal technology represent an emerging intersection between civil law, artificial intelligence, compliance, evidence and legal ethics.

The UAE's codified legal structure provides an important framework for controlling these systems. AI can:

• identify patterns;
• organize evidence;
• assess risk;
• detect potential abuse;
• recommend proportionate responses;
• support compliance;
• assist experts; and
• improve legal efficiency.

But AI should not independently determine:

• liability;
• ownership;
• fraud;
• bad faith;
• abuse of rights;
• criminal responsibility;
• enforceability of a contract; or
• the final legal rights of parties.

The strongest legal model is therefore:

Law → Evidence → AI analysis → Ethical/proportionality assessment → Human review → Legally authorized decision → Judicial accountability.

The UAE case law concerning evidence, expert reports, good faith, abuse of rights, contractual powers and AML demonstrates an underlying principle that remains highly relevant to AI: legal decisions must ultimately be grounded in legally relevant facts, reliable evidence, reasoned evaluation and lawful authority.

The pre-2026 cases discussed above should be treated principally as foundational or analogical authorities, because they were decided under the earlier legal framework. The current analysis must also take account of the Civil Transactions Law effective from 1 June 2026, the current UAE Evidence Law, and the 2025 AML framework.